TinySR: Pruning Diffusion for Real-World Image Super-Resolution

2025-08-24

AI Summary

To address the challenge of balancing complex degradations and efficient inference in real-world image super-resolution (Real-ISR), this paper proposes TinySR—a lightweight diffusion-based model. TinySR employs a suite of synergistic optimizations: dynamic inter-block activation control, expansion-erosion channel pruning, separable convolution replacement, VAE latent-space compression, removal of attention and temporal/prompt modules, and pre-caching of critical features—thereby drastically reducing computational overhead while preserving generative priors. Compared to the teacher model TSD-SR, TinySR achieves up to 5.68× inference speedup and an 83% reduction in parameter count, while maintaining competitive performance in PSNR, LPIPS, and perceptual quality. To the best of our knowledge, TinySR is the first diffusion-based method to enable real-time, high-fidelity Real-ISR restoration.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Real-world image super-resolution (Real-ISR) focuses on recovering high-quality images from low-resolution inputs that suffer from complex degradations like noise, blur, and compression. Recently, diffusion models (DMs) have shown great potential in this area by leveraging strong generative priors to restore fine details. However, their iterative denoising process incurs high computational overhead, posing challenges for real-time applications. Although one-step distillation methods, such as OSEDiff and TSD-SR, offer faster inference, they remain fundamentally constrained by their large, over-parameterized model architectures. In this work, we present TinySR, a compact yet effective diffusion model specifically designed for Real-ISR that achieves real-time performance while maintaining perceptual quality. We introduce a Dynamic Inter-block Activation and an Expansion-Corrosion Strategy to facilitate more effective decision-making in depth pruning. We achieve VAE compression through channel pruning, attention removal and lightweight SepConv. We eliminate time- and prompt-related modules and perform pre-caching techniques to further speed up the model. TinySR significantly reduces computational cost and model size, achieving up to 5.68x speedup and 83% parameter reduction compared to its teacher TSD-SR, while still providing high quality results.

Problem

• Reducing computational overhead in diffusion models for super-resolution
• Addressing large over-parameterized model architectures in Real-ISR
• Achieving real-time performance while maintaining perceptual quality

Innovation

• Dynamic Inter-block Activation for depth pruning
• VAE compression via channel pruning and attention removal
• Pre-caching techniques to eliminate time-related modules